nltk-users Google Group

Tagging phrases with type

2013-06-18T06:31:41Z

Hi,
I am chunking certain phrases and then marking type before POS tagging
them. For example, "windows xp'' as OS. I have a corpus where phrases are
classified as per their type. I want an efficient nltk tool to perform this
task. I looked at parsers but context free grammars require complete
coverage which I cant provide because there will be some terminals in my

Parsing, Feature Structures, and Semantics

2013-06-17T22:54:20Z

I am trying to extract semantics from sentences, and using Python & NLTK as
a learning toolkit.

If I understand the (Bird et al) book and the NLTK Semantics HowTo, first
order semantics are extracted from a parse tree using root_semrep().
However, root_semrep() requires a parse tree with feature structures.

Re: parsing issues on -ing and -ed words

2013-06-17T22:36:59Z

Hi Gene,
You can override tags from the default tagger by creating your own tagger,
as described
in [link]
Or, you can try training your own pos tagger:
[link]

Re: phrasal replacement?

2013-06-17T22:32:11Z

Hi Gene,
The CsvWordReplacer & RegexReplacer are examples from my cookbook, so
you're welcome to modify them however you want. If you do make useful
modifications, please share them - I'd happily post them on
[link] or maybe include them in a second edition of the
cookbook (with your permission).

problem with LazyCorpusLoader

2013-06-16T12:23:18Z

i'm using nltk.corpus.util.LazyCorpusLoa der() to load wsj treebank mrg
files with tag_mapping_function=nltk.tag. simplify argument but even
after doing so i'm getting non terminals like 'NP-SBJ-1'

this is the line of code i'm using to load the treebank
treebank=nltk.corpus.util.Lazy CorpusLoader('treebank/

Discussing Exercise 24 and 27 of Ch02 of the Book

2013-06-15T08:19:09Z

Hello, everyone, I started reading the book, NLP with Python, recently, and
have just finished Chap 02 last night. In completing this chapter, I find
Exercise 24 very interesting and would love to discuss with you guys.
Exercise 24 approaches the random text generation function. A primitive
version of this function, which, given the seed word, chooses every next

phrasal replacement?

2013-06-14T01:31:29Z

I wanted to use the CsvWordReplacer to create a phrasal replacement
file for things such as idioms and phrasal prepositions.

Examples:
kicked the bucket >> kicked_the_bucket
because of >> because_of
in fact >> in_fact

I'm running into two issues with this.
CsvWordReplacer expects single tokens, not multiples and/or sentences.

parsing issues on -ing and -ed words

2013-06-14T01:15:02Z

I'm noticing some rather odd parsing results in NLTK that I'd like to
find a way to correct. I would welcome any suggestions anyone might
have, especially if you've encountered this issue and have found a
solution.

Try the following sentence:
"The pyramids, for example, bring the eye to a point, and that point

categorized corpus

2013-06-09T12:04:13Z

Hi,
my situation is just like movie_review corpus
my folder name is commenFb, so inside this folder have 2 folder, i label it
one is positif another negatif.
The problem is when i call reader.categories(),there is nothing display.
So, how i want to make this categories call my folder.
this my coding.

Re: [nltk-users] Re: create my own corpus

2013-06-06T19:27:01Z

You are welcome.

Your corpus is ready to use, it's just loaded differently from the
NLTK's built-in corpora. If you look inside the source for nltk.corpus,
you'll see something like this (simplifying for exposition):

brown = CategorizedTaggedCorpusReader( ...)

It's just a trick, in other words. When you import brown, you import a

Re: create my own corpus

2013-06-06T06:28:05Z

the PlaintextCorpusReader is work.
from nltk.corpus.reader import PlaintextCorpusReader
os.path.join('C:\\Users\\Win 7\\nltk_data', "corpora/muet")
corpus_root = 'C:\\Users\\Win 7\\nltk_data\\corpora/muet'
wordlists = PlaintextCorpusReader(corpus_r oot,'.*')
wordlists.fileids()
when i want to import the muet such as (from nltk.corpus import muet) to

Re: [nltk-users] create my own corpus

2013-06-05T12:04:49Z

Nana,

Take a loot at the documentation for the PlaintextCorpusReader class.
You call it with the (relative) path to your corpus, e.g.,
os.path.join(nltk.data.root, "corpora/muet"), and a pattern matching the
filenames, and you get a corpus object with methodsfileids(), words(),
sents(), etc.

create my own corpus

2013-06-05T06:46:36Z

I want to create my own corpus.
I have my own folder document which i label it as "muet".
I already copy this folder into the path where corpora place.
In this folder, i have more than one file text.
So, how i want to load this folder into the NLTK, so that NLTK can access
my file.

Named Entity Recognition with lower case words?

2013-06-01T07:12:30Z

Hi all,
I am developing a software which analyze the social media and come with new
trends and I am using nltk to achieve my goals.
My question is In nltk when we use the NER feature, if we input
the sentence "*I live in New York*." , it will identify "New York" as a *
Location*.
But if the input is "*I live in new york*." it wont recognize "new york" as

Question on Tagging

2013-05-30T19:34:02Z

Dear Group,
From web surfing I came to know the simple "nltk.pos_tag(text)" is based on
Maximum Entropy Model.
Is the information fine?
Regards,
Subhabrata.